Model-based control of cardiac alternans on a ring

Optical Ultrastructure of Large Mammalian Hearts Recovers Discordant Alternans by In Silico Data Assimilation

Understanding and predicting the mechanisms promoting the onset and sustainability of cardiac arrhythmias represent a primary concern in the scientific and medical communities still today. Despite the long-lasting effort in clinical and physico-mathematical research, a critical aspect to be fully characterized and unveiled is represented by spatiotemporal alternans patterns of cardiac excitation. The identification of discordant alternans and higher-order alternating rhythms by advanced data analyses as well as their prediction by reliable mathematical models represents a major avenue of research for a broad and multidisciplinary scientific community. Current limitations concern two primary aspects: 1) robust and general-purpose feature extraction techniques and 2) in silico data assimilation within reliable and predictive mathematical models. Here, we address both aspects. At first, we extend our previous works on Fourier transformation imaging (FFI), applying the technique to whole-ventricle fluorescence optical mapping. Overall, we identify complex spatial patterns of voltage alternans and characterize higher-order rhythms by a frequency-series analysis. Then, we integrate the optical ultrastructure obtained by FFI analysis within a fine-tuned electrophysiological mathematical model of the cardiac action potential. We build up a novel data assimilation procedure demonstrating its reliability in reproducing complex alternans patterns in two-dimensional computational domains. Finally, we prove that the FFI approach applied to both experimental and simulated signals recovers the same information, thus closing the loop between the experiment, data analysis, and numerical simulations.

Long-Time Prediction of Arrhythmic Cardiac Action Potentials Using Recurrent Neural Networks and Reservoir Computing

The electrical signals triggering the heart's contraction are governed by non-linear processes that can produce complex irregular activity, especially during or preceding the onset of cardiac arrhythmias. Forecasts of cardiac voltage time series in such conditions could allow new opportunities for intervention and control but would require efficient computation of highly accurate predictions. Although machine-learning (ML) approaches hold promise for delivering such results, non-linear time-series forecasting poses significant challenges. In this manuscript, we study the performance of two recurrent neural network (RNN) approaches along with echo state networks (ESNs) from the reservoir computing (RC) paradigm in predicting cardiac voltage data in terms of accuracy, efficiency, and robustness. We show that these ML time-series prediction methods can forecast synthetic and experimental cardiac action potentials for at least 15–20 beats with a high degree of accuracy, with ESNs typically two orders of magnitude faster than RNN approaches for the same network size.

Controllability of voltage- and calcium-driven cardiac alternans in a map model

Certain cardiac arrhythmias are preceded by electrical alternans, a state characterized by beat-to-beat alternation in cellular action potential duration. Cardiac alternans may arise from different mechanisms including instabilities in voltage or intracellular calcium cycling. Although a number of techniques have been proposed to suppress alternans, these methods have mainly been tested using models that do not support calcium-driven alternans. Therefore, it is important to understand how control methods may perform when alternans is driven by instabilities in calcium cycling. In this study, we applied controllability analysis to a discrete map of alternans dynamics in a cardiac cell. We compared two different controllability measures to determine to what extent different control strategies could suppress alternans and tested these predictions using three feedback controllers. We found a modal controllability measure, unlike the minimum singular value of the controllability matrix, consistently indicated the control strategies requiring the least control effort and yielding the smallest closed-loop eigenvalue. In addition, action potential duration was identified as the most effective variable through which control can be applied, regardless of alternans mechanism, although sarcoplasmic reticulum calcium load was also useful for the calcium-driven alternans cases.

Global vs local control of cardiac alternans in a 1D numerical model of human ventricular tissue

Cardiac alternans is a proarrhythmic state in which the action potential duration (APD) of cardiac myocytes alternate between long and short values and often occurs under conditions of rapid pacing of cardiac tissue. In the ventricles, alternans is especially dangerous due to the life-threatening risk of developing arrhythmias, such as ventricular fibrillation. Alternans can be formed in periodically paced tissue as a result of pacing itself. Recently, it has been demonstrated that this pacing-induced alternans can be prevented by performing constant diastolic interval (DI) pacing, in which DI is independent of APD. However, constant DI pacing is difficult to implement in experimental settings since it requires the real-time measurement of APD. A more practical way was proposed based on electrocardiograms (ECGs), which give an indirect measure of the global DI relaxation period through the TR interval assessment. Previously, we demonstrated that constant TR pacing prevented alternans formation in isolated Langendorff-perfused rabbit hearts. However, the efficacy of "local" constant DI pacing vs "global" constant TR pacing in preventing alternans formation has never been investigated. Thus, the purpose of this study was to implement an ECG-based constant TR pacing in a 1D numerical model of human ventricular tissue and to compare the dynamical behavior of cardiac tissue with that resulted from a constant DI pacing. The results showed that both constant TR and constant DI pacing prevented the onset of alternans until lower basic cycle length when compared to periodic pacing. For longer cable lengths, constant TR pacing was shown to exhibit greater control on alternans than constant DI pacing.

Simulating waves, chaos and synchronization with a microcontroller

The spatiotemporal dynamics of complex systems have been studied traditionally and visualized numerically using high-end computers. However, due to advances in microcontrollers, it is now possible to run what once were considered large-scale simulations using a very small and inexpensive single integrated circuit that can furthermore send and receive information to and from the outside world in real time. In this paper, we show how microcontrollers can be used to perform simulations of nonlinear ordinary differential equations with spatial coupling and to visualize their dynamics using arrays of light-emitting diodes and/or touchscreens. We demonstrate these abilities using three different models: two reaction-diffusion models (one neural and one cardiac) and a generic model of network oscillators. These models are commonly used to simulate various phenomena in biophysical systems, including bifurcations, waves, chaos, and synchronization. We also demonstrate how simple it is to integrate real-time user interaction with the simulations by showing examples with a light sensor, touchscreen, and web browser.

Real-Time Closed Loop Diastolic Interval Control Prevents Cardiac Alternans in Isolated Whole Rabbit Hearts

Cardiac alternans, a beat-to-beat alternation in action potential duration (APD), can lead to fatal arrhythmias. During periodic pacing, changes in diastolic interval (DI) depend on subsequent changes in APD, thus enhancing cardiac instabilities through a 'feedback' mechanism. Recently, an anti-arrhythmic Constant DI pacing protocol was proposed and shown to be effective in suppressing alternans in 0D and 1D in silico studies. However, previous experimental validation of Constant DI pacing in the heart has been unsuccessful due to the spatio-temporal complexity of 2D cardiac tissue and the technical challenges in its real-time implementation. Here, we developed a novel closed loop system to detect T-waves from real-time ECG data, enabling successful implementation of Constant DI pacing protocol, and performed high-resolution optical mapping experiments on isolated whole rabbit hearts to validate its anti-arrhythmic effects. The results were compared with: (1) Periodic pacing (feedback inherent) and (2) pacing with heart rate variability (HRV) (feedback modulation) introduced by using either Gaussian or Physiological patterns. We observed that Constant DI pacing significantly suppressed alternans in the heart, while maintaining APD spatial dispersion and flattening the slope of the APD restitution curve, compared to traditional Periodic pacing. In addition, introduction of HRV in Periodic pacing failed to prevent cardiac alternans, and was arrhythmogenic.

Theory of the development of alternans in the heart during controlled diastolic interval pacing

The beat-to-beat alternation in action potential durations (APDs) in the heart, called APD alternans, has been linked to the development of serious cardiac rhythm disorders, including ventricular tachycardia and fibrillation. The length of the period between action potentials, called the diastolic interval (DI), is a key dynamical variable in the standard theory of alternans development. Thus, methods that control the DI may be useful in preventing dangerous cardiac rhythms. In this study, we examine the dynamics of alternans during controlled-DI pacing using a series of single-cell and one-dimensional (1D) fiber models of alternans dynamics. We find that a model that combines a so-called memory model with a calcium cycling model can reasonably explain two key experimental results: the possibility of alternans during constant-DI pacing and the phase lag of APDs behind DIs during sinusoidal-DI pacing. We also find that these results can be replicated by incorporating the memory model into an amplitude equation description of a 1D fiber. The 1D fiber result is potentially concerning because it seems to suggest that constant-DI control of alternans can only be effective over only a limited region in space.

Nonlinear physics of electrical wave propagation in the heart: a review

The beating of the heart is a synchronized contraction of muscle cells (myocytes) that is triggered by a periodic sequence of electrical waves (action potentials) originating in the sino-atrial node and propagating over the atria and the ventricles. Cardiac arrhythmias like atrial and ventricular fibrillation (AF,VF) or ventricular tachycardia (VT) are caused by disruptions and instabilities of these electrical excitations, that lead to the emergence of rotating waves (VT) and turbulent wave patterns (AF,VF). Numerous simulation and experimental studies during the last 20 years have addressed these topics. In this review we focus on the nonlinear dynamics of wave propagation in the heart with an emphasis on the theory of pulses, spirals and scroll waves and their instabilities in excitable media with applications to cardiac modeling. After an introduction into electrophysiological models for action potential propagation, the modeling and analysis of spatiotemporal alternans, spiral and scroll meandering, spiral breakup and scroll wave instabilities like negative line tension and sproing are reviewed in depth and discussed with emphasis on their impact for cardiac arrhythmias.

Continuous-time control of alternans in long Purkinje fibers

Alternans-an arrhythmic response of cardiac tissue to periodic pacing-often serves as a precursor to a more dangerous, and potentially lethal, state of fibrillation. Suppression of alternans using feedback control may be a plausible method to prevent fibrillation. Several approaches based on impulsive control have been proposed previously, where feedback is applied for a brief instance of time during each pacing interval. This paper presents a continuous-time approach, where feedback current is applied at all times, which is capable of suppressing alternans in fibers of significantly greater length (up to at least 4 cm), compared with impulsive control (less than 1 cm), and for a wide range of pacing cycle lengths.

Control of cardiac alternans by mechanical and electrical feedback

A persistent alternation in the cardiac action potential duration has been linked to the onset of ventricular arrhythmia, which may lead to sudden cardiac death. A coupling between these cardiac alternans and the intracellular calcium dynamics has also been identified in previous studies. In this paper, the system of PDEs describing the small amplitude of alternans and the alternation of peak intracellular Ca(2+) are stabilized by optimal boundary and spatially distributed actuation. A simulation study demonstrating the successful annihilation of both alternans on a one-dimensional cable of cardiac cells by utilizing the full-state feedback controller is presented. Complimentary to these studies, a three variable Nash-Panfilov model is used to investigate alternans annihilation via mechanical (or stretch) perturbations. The coupled model includes the active stress which defines the mechanical properties of the tissue and is utilized in the feedback algorithm as an independent input from the pacing based controller realization in alternans annihilation. Simulation studies of both control methods demonstrate that the proposed methods can successfully annihilate alternans in cables that are significantly longer than 1 cm, thus overcoming the limitations of earlier control efforts.

Shock-induced termination of reentrant cardiac arrhythmias: comparing monophasic and biphasic shock protocols

In this article, we compare quantitatively the efficiency of three different protocols commonly used in commercial defibrillators. These are based on monophasic and both symmetric and asymmetric biphasic shocks. A numerical one-dimensional model of cardiac tissue using the bidomain formulation is used in order to test the different protocols. In particular, we performed a total of 4.8 × 10(6) simulations by varying shock waveform, shock energy, initial conditions, and heterogeneity in internal electrical conductivity. Whenever the shock successfully removed the reentrant dynamics in the tissue, we classified the mechanism. The analysis of the numerical data shows that biphasic shocks are significantly more efficient (by about 25%) than the corresponding monophasic ones. We determine that the increase in efficiency of the biphasic shocks can be explained by the higher proportion of newly excited tissue through the mechanism of direct activation.

Model-based control of cardiac alternans in Purkinje fibers

This paper describes a systematic approach to suppressing cardiac alternans in simulated Purkinje fibers using localized current injections. We investigate the controllability and observability of the periodically paced Noble model for different locations of the recording and control electrodes. In particular, we show that the loss of controllability causes the failure of the control approach introduced by Echebarria and Karma [Chaos 12, 923 (2002)] for longer fiber lengths. Furthermore, we explain how the optimal locations for the recording and control electrodes and the timing of the feedback current can be selected, accounting for both linear and nonlinear effects, effectively doubling the length of fibers that can be controlled with previous methods.

Termination of reentrant cardiac action potential propagation using far-field electrical pacing

Several different types of rapid cardiac rhythm disorders, including atrial and ventricular fibrillation, are likely caused by multiple, rapidly rotating, action potential (AP) waves. Thus, an electrical pacing therapy, whose effectiveness is based on being delivered with a particular timing relative to one of these waves, is unlikely to be useful in terminating the remaining waves. Here, we develop pacing protocols that are designed to terminate rotating waves independently of when the sequences of stimuli are imposed or where each wave is in its rotation at the time the sequences are initiated. These protocols are delivered as far-field stimuli, and therefore are capable of simultaneously influencing all the waves present. The pacing intervals for these protocols are, in general, of unequal duration and are determined through examination of the dynamics of AP propagation in a 1-D ring model. Series of two or three stimuli with interstimulus intervals chosen in this way are shown to be effective in terminating these waves over a wide range of ring circumferences and AP dynamical parameters. Stimulus sequences of this type may form the basis for developing new defibrillation protocols to test in experiments or more realistic models of the electrical heart.

Applications of control theory to the dynamics and propagation of cardiac action potentials

Sudden cardiac arrest is a widespread cause of death in the industrialized world. Most cases of sudden cardiac arrest are due to ventricular fibrillation (VF), a lethal cardiac arrhythmia. Electrophysiological abnormalities such as alternans (a beat-to-beat alternation in action potential duration) and conduction block have been suspected to contribute to the onset of VF. This study focuses on the use of control-systems techniques to analyze and design methods for suppressing these precursor factors. Control-systems tools, specifically controllability analysis and Lyapunov stability methods, were applied to a two-variable Karma model of the action-potential (AP) dynamics of a single cell, to analyze the effectiveness of strategies for suppressing AP abnormalities. State-feedback-integral (SFI) control was then applied to a Purkinje fiber simulated with the Karma model, where only one stimulating electrode was used to affect the system. SFI control converted both discordant alternans and 2:1 conduction block back toward more normal patterns, over a wider range of fiber lengths and pacing intervals compared with a Pyragas-type chaos controller. The advantages conferred by using feedback from multiple locations in the fiber, and using integral (i.e., memory) terms in the controller, are discussed.